Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 2172

Alphabet chairman Eric Schmidt says the US is at risk of falling behind in the race to develop cutting-edge artificial intelligence. Speaking at a tech summit organized by national security think tank CNAS, Schmidt predicted that America’s lead in the field would continue “over the next five years” before China catches up “extremely quickly.”

“They are going to use this technology for both commercial and military objectives, with all sorts of implications,” said Schmidt, referencing a Chinese policy document outlining the country’s ambition to become the global leader in AI by 2030. Schmidt reiterated several familiar talking points in this debate, primarily that the US is failing to invest in basic research, and that a restrictive immigration policy hobbles the country’s ability to attract AI talent from overseas.

“Some of the very best people are in countries that we won’t let into America. Would you rather have them building AI somewhere else, or rather have them here?” said Schmidt. “Iran produces some of the top computer scientists in the world, and I want them here. To be clear, I want them working for Alphabet and Google!”

Read more

Artificial intelligence beats over 100 London lawyers in predicting case outcome:

In a contest that took place last month. It pitched over 100 lawyers from many of London’s ritziest firms against an artificial intelligence program called Case Cruncher Alpha.

Both the humans and the AI were given the basic facts of hundreds of PPI (payment protection insurance) mis-selling cases and asked to predict whether the Financial Ombudsman would allow a claim.

Read more

After more than a decade away, Sony’s Aibo pet robot is making a return. The original dog-like robot launched in 1999, while Sony says its followup is “capable of forming an emotional bond with members of the household while providing them with love, affection, and the joy of nurturing and raising a companion.” Its OLED eyes allow for “nuanced” expressions as fisheye cameras see and recognize individual faces while new actuators allow its body to move smoothly along 22 axes. Pre-orders for the new robot begin tonight via Sony’s online store in Japan for 179,000 yen (about $1,739 US), with shipments scheduled to begin on January 11th.

Of course, because this is 2017, not only is the new Aibo powered by AI (that learns and develops a unique personality over time) but it’s also connected to the cloud. An Aibo Basic Plan subscription not only backs up your robot’s unique identity but also turns on the connection for owners to access their remote via WiFi or a mobile connection.

You’ll need the subscription for your pet’s AI tendencies to develop, and use its accompanying My Aibo app (on iOS, Android and the web) to manage settings, access photos it takes and even play with a virtual version of the dog when you’re away. Eventually, you’ll be able to buy new tricks from the Aibo Store, but it’s launching with one accessory — the Aibone.

Read more

Robots have become central to the cleaning-up operation at Japan’s Fukushima nuclear power plant, six years after the tsunami that triggered the nuclear meltdown.

It is estimated that around 600 tonnes of toxic fuel may have leaked out of the reactor during the incident.

The Tokyo Electric Power Company is using a variety of robots to explore areas too dangerous for people to go near.

Read more

For any AI people out there. I’d really like to see an AI get dropped into Ocarina of Time, and then Skyrim. The day an AI can be dropped into those, and complete the entire games, and go out and complete all the weird random tasks, it should be pretty close to human level.

AI just beat a top human professional in the game Dota 2, but the technology could help with much bigger strategic problems.

Read more

Thank a new approach to spoofing image recognition AIs, developed by a team from Kyushu University in Japan, for that joke.

Trying to catch out AIs is a popular pastime for many researchers, and we’ve reported machine-learning spoofs in the past. The general approach is to add features to images that will incorrectly trigger a neural network and have it identify what it sees as something else entirely.

The new research, published on the arXiv, describes an algorithm that can efficiently identify the best pixels to alter in order to confuse an AI into mislabeling a picture. By changing just one pixel in a 1,024-pixel image, the software can trick an AI about 74 percent of the time. That figure rises to around 87 percent if five pixels are tweaked.

Read more